面向对象的遥感震害信息提取方法——以汶川地震为例  

Object-oriented Method for Extracting Seismic Disaster Information Using Remote Sensing Images --A Case Study of Wenchuan Earthquake

作　　者：赵福军[1,2] 张磊[3] 

机构地区：[1]中国地震局工程力学研究所,黑龙江哈尔滨150080 [2]黑龙江科技学院,黑龙江哈尔滨150027 [3]天津市地震局,天津300201

出　　处：《地震》2009年第B10期130-138,共9页Earthquake

基　　金：“十一五”同家科技支撑计划（2006BAC13B03-0102）

摘　　要：本文总结了近年来利用遥感影像进行震害信息提取的方法和研究趋势，随着遥感影像分辨率的提高，震害信息提取逐渐从传统的基于像元方法向基于对象的方法转变。文中以汶川地震为例，利用面向对象的分类方法提取了典型震害信息，并提出了面向对象的变化检测方法的工作流程。对震害建筑物破坏情况，分别用最小距离分类、马氏距离分类、支撑向量机分类和面向对象分类进行了实验研究，并对这几种不同分类方法的实验结果进行了对比分析。实验结果表明，面向对象的震害信息提取方法克服了传统的基于像元的分类方法的缺点，提高了建筑物识别的精度，在建筑物震害和地震次生灾害的信息提取中取得了较好的效果。This paper summarizes the method and research trends on extracting seismic disaster information using remote sensing images in recent years. With the increase of remote sensing images＇ resolution, the extraction of seismic disaster information changes gradually from pixel-based method to object-based method. In a case study of the Wenchuan earthquake, this paper utilizes object-oriented classification method in extracting typical seismic disaster information, and presents the workflow of object-oriented change detection method. Experiments on seismic disaster building damage were conducted using minimum distance classification, Mahalanobis distance classification, SVM classification, object-oriented classification methods. Comparative analysis was carried out on the corresponding results of different classification method. The experimental results show that object-oriented seismic disaster information extraction method can overcome the disadvantage of conventional pixel-oriented method, improve the building identification precision, and get good effects in extracting building seismic disaster and secondary seismic disaster information.

关 键 词：多尺度分割 面向对象变化检测 遥感 震害信息提取 

分 类 号：P315.7[天文地球—地震学]